Jurisdictions with effective smoking bans large enough to be analyzed in the data, 2000–2005

Note: Only counties for which we have sufficient data for 2000–2005 are analyzed in the sample. Any bans enacted in the latter half of 2005 are not included in the analysis and the information from the county is dropped. Several counties (e.g., Suffolk County, New York and Suffolk County, Massachusetts) are included in the treatment group as soon as they pass the ban, with the remainder of the state added later. A total of 117 counties compose the treatment group. Benton County, OR and California passed bans prior to 2000. Given they provide us with no identifying information, they are not included above.

Means and proportions of key variables in analysis

Note: Since logs are used for the main specifications, we only calculate the means for counties with positive numbers of fatal accidents for a given year. We verify that this does not bias the results. All estimates are weighted by county-year population.

Determinants of fatal accidents involving alcohol, 2000–2005

Note: Each column is from a separate weighted least squares regression that includes both county and year fixed effects. The dependent variable is the log of the estimated number of fatal accidents involving a driver with a BAC exceeding 0.08 in a county-year. The standard errors in parentheses are corrected to allow for non-independence of observations within a county through clustering. The fifth column excludes bans passed in the college towns of Lawrence (Kansas), Lexington (Kentucky), and Lincoln (Nebraska) that did not cover the rest of the county. The final column (6) excludes the 23 (of 117) counties for which a ban became effective after July 1, 2004 (Massachusetts through Tift County in Table 1). The remaining bans therefore have been in place for at least 18 months and two years of post-ban alcohol-related accidents are reliably identified.

Effects of smoking bans over time

Note: Each column is from a separate weighted least squares regression. The dependent variable is the log of the estimated number of fatal accidents involving a driver with a BAC exceeding 0.08. County and year fixed effects, as well as controls for accidents not involving alcohol, population, beer tax level, and minimum BAC levels, are included in all regressions. The standard errors in parentheses are corrected to allow for non-independence of observations within a county through clustering.

Robustness of the basic results

Note: Each cell is from a separate regression. County and year fixed effects, as well as controls for accidents not involving alcohol, population, beer tax, and minimum BAC levels, are included in all regressions. In the robustness checks using alternative control groups, the sample size will differ. We note the number of counties used in parentheses in these cases. Coefficient estimates and standard errors (corrected to allow for non-independence of observations within a county through clustering unless otherwise noted) are reported.

Effect of smoking bans in places more and less prone to drunk driving

Note: Each row is from a separate weighted least squares regression. The dependent variable is the log of the estimated number of fatal accidents involving a driver with a BAC exceeding 0.08. County and year fixed effects, as well as controls for accidents not involving alcohol, population, beer tax level, and minimum BAC levels, are included in all regressions. The standard errors in parentheses are corrected to allow for non-independence of observations within a county through clustering.

Border county case studies

Note: The dependent variable is the log of the estimated number of fatal accidents involving a driver with a BAC exceeding 0.08. County and year fixed effects, as well as controls for accidents not involving alcohol, population, beer tax level, and minimum BAC levels, are included in all regressions. The standard errors in parentheses are corrected to allow for non-independence of observations within a county through clustering. The control group consists of all counties that do not have a smoke-free ordinance and are not a border county.

Additional case studies; within jurisdiction effects

Note: The dependent variable is the log of the estimated number of fatal accidents involving a driver with a BAC exceeding 0.08. County and year fixed effects, as well as controls for accidents not involving alcohol, population, beer tax level, and minimum BAC levels, are included in all regressions. The standard errors in parentheses are corrected to allow for non-independence of observations within a county through clustering.